[MEP]

For the projects where each step is pretty clear in advance, I just break the project down as much as possible into atomic little actions. I don't feel as though I've wasted time planning when doing this because it really does give me a good idea of what the project entails and what kind of time commitment it represents. It also allows me to make progress on that project bit by bit rather than waiting to do several things before checking that "complete" box on a task that wasn't broken down as far as it could be.

For those projects that evolve over time, I take my best guess at what the steps should look like and make changes as I go along if I need to. When I complete the current action, I see what action becomes available next and decide if it makes any sense in light of what I did in the last action. If not, I change it right then, and then I take a look at the whole project and do a quick review of it alone.

You could do that in the context of a larger review too, but I find it more helpful to make those changes while the knowledge gained from the last step is still fresh in my mind. This takes, at most, a couple minutes for me so I don't feel I'm wasting too much time refactoring in these situations.